Alibaba Group Holding announced that its open-weight Qwen family of artificial-intelligence models has been downloaded in excess of 3 billion instances worldwide over the past half-year. The figure eclipses the 418 million downloads recorded for Google and the 227 million for Meta Platforms Inc., according to a recent state-of-open-models report from the AI hub Hugging Face.
Record download figures
The report, published on 14 August, notes that Alibaba has open-sourced more than 460 individual models and that its ecosystem now supports over 300 000 derivative versions. By contrast, the leading US firms have far fewer downloads, underscoring the rapid uptake of Qwen among developers seeking adaptable, low-cost foundations for new AI products.
Why the surge matters
Open-source models can be freely downloaded, customised and incorporated into downstream applications, making download counts a proxy for influence in the ongoing US-China AI competition. The rise of Qwen suggests that Chinese developers are successfully positioning their technology as a viable alternative to closed-source offerings from companies such as OpenAI Inc. and Anthropic PBC. Export controls on chips and AI systems have not slowed this momentum, as evidenced by the continued growth of Chinese-origin models like Qwen, Moonshot AI Inc., and DeepSeek.
Qwen has become part of the default workflow for developers deciding what models to fine-tune and deploy,
The report adds that the breadth of the Qwen family creates a self-reinforcing cycle: broader adoption fuels more derivative models, which in turn attracts additional users. Alibaba has amplified this effect by offering Qwen through its cloud platform to enterprise customers across Southeast Asia and Africa, extending its reach beyond domestic rivals such as DeepSeek, Moonshot Kimi and MiniMax.
Future outlook
US technology giants are responding. In recent weeks, Meta and Nvidia Corp. have launched new open-source AI models to compete for developer attention. The next phase will likely see intensified rivalry as both sides accelerate model releases and improve accessibility, with the ultimate impact on the global AI ecosystem still unfolding.